[feature] add support for dft loss (#8917)

This commit is contained in:
XLXW 2025-08-15 23:29:57 +08:00 committed by GitHub
parent 2b66b4df43
commit 0454c10456
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
4 changed files with 97 additions and 0 deletions

View File

@ -0,0 +1,43 @@
### model
model_name_or_path: Qwen/Qwen2-1.5B-Instruct
trust_remote_code: true
### method
stage: sft
do_train: true
finetuning_type: full
use_dft_loss: true
### dataset
dataset: identity,alpaca_en_demo
template: qwen
cutoff_len: 2048
max_samples: 1000
overwrite_cache: true
preprocessing_num_workers: 16
dataloader_num_workers: 4
### output
output_dir: saves/qwen2-1_5b/full/sft
logging_steps: 10
save_steps: 500
plot_loss: true
overwrite_output_dir: true
save_only_model: false
report_to: none # choices: [none, wandb, tensorboard, swanlab, mlflow]
### train
per_device_train_batch_size: 1
gradient_accumulation_steps: 8
learning_rate: 1.0e-5
num_train_epochs: 3.0
lr_scheduler_type: cosine
warmup_ratio: 0.1
bf16: true
ddp_timeout: 180000000
### eval
# val_size: 0.1
# per_device_eval_batch_size: 1
# eval_strategy: steps
# eval_steps: 500

View File

@ -428,6 +428,10 @@ class FinetuningArguments(
default=False,
metadata={"help": "Whether or not to use the Muon optimizer."},
)
use_dft_loss: bool = field(
default=False,
metadata={"help": "Whether to use the DFT loss."},
)
freeze_vision_tower: bool = field(
default=True,
metadata={"help": "Whether ot not to freeze the vision tower in MLLM training."},

View File

@ -78,6 +78,11 @@ class CustomSeq2SeqTrainer(Seq2SeqTrainer):
self.accelerator.clip_grad_norm_ = MethodType(clip_grad_norm_old_version, self.accelerator)
self.add_callback(BAdamCallback)
if finetuning_args.use_dft_loss:
from ..trainer_utils import dft_loss_func
self.compute_loss_func = dft_loss_func
@override
def create_optimizer(self) -> "torch.optim.Optimizer":
if self.optimizer is None:

View File

@ -631,6 +631,51 @@ def get_batch_logps(
return logps, valid_length
def dft_loss_func(outputs, labels, num_items_in_batch=None):
logits = outputs.get("logits")
if logits is None:
return outputs.get("loss", torch.tensor(0.0))
logits = logits.float()
vocab_size = logits.size(-1)
labels = torch.nn.functional.pad(labels, (0, 1), value=-100)
shift_labels = labels[..., 1:].contiguous()
logits = logits.view(-1, vocab_size)
shift_labels = shift_labels.view(-1)
shift_labels = shift_labels.to(logits.device)
loss = _dft_cross_entropy(logits, shift_labels, num_items_in_batch)
return loss
def _dft_cross_entropy(
source: torch.Tensor,
target: torch.Tensor,
num_items_in_batch: Optional[torch.Tensor] = None,
ignore_index: int = -100,
) -> torch.Tensor:
per_token_loss = torch.nn.functional.cross_entropy(source, target, ignore_index=ignore_index, reduction="none")
valid_mask = target != ignore_index
if not valid_mask.any():
return torch.tensor(0.0, device=source.device, dtype=source.dtype)
valid_losses = per_token_loss[valid_mask]
with torch.no_grad():
target_probs = torch.exp(-valid_losses)
weighted_losses = valid_losses * target_probs
if num_items_in_batch is not None:
total_loss = weighted_losses.sum()
if torch.is_tensor(num_items_in_batch):
num_items_in_batch = num_items_in_batch.to(total_loss.device)
loss = total_loss / num_items_in_batch
else:
loss = weighted_losses.mean()
return loss
def nested_detach(
tensors: Union["torch.Tensor", list["torch.Tensor"], tuple["torch.Tensor"], dict[str, "torch.Tensor"]],
clone: bool = False,